🚀 The feature, motivation and pitch
The scaled_mm(pytorch/pytorch#165978) and inference(pytorch/ao#3248) is WIP, You can enable the realted training UTs with emulate path and comments your PR in pytorch/ao#2917 to help us to track the total status.
Reference: https://github.com/pytorch/ao/tree/main/torchao/prototype/mx_formats
Alternatives
No response
Additional context
No response